Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper introduces the Bayesian X-Learner, a meta-learner for estimating heterogeneous treatment effects that combines cross-fitted doubly robust pseudo-outcomes with a full MCMC posterior inference using a Welsch redescending pseudo-likelihood to provide calibrated uncertainty. It addresses the challenge of estimating Conditional Average Treatment Effects (CATE) under heavy-tailed outcome distributions, where existing methods often fall short in providing both accurate effect estimates and well-calibrated uncertainty. Experiments on the IHDP benchmark and contaminated "whale" DGPs demonstrate competitive performance and robustness to heavy-tailed data, achieving good RMSE and credible interval coverage.
Finally, a CATE estimator that delivers both heterogeneous treatment effects and calibrated uncertainty even when your outcome data is riddled with heavy tails.
Conditional Average Treatment Effect (CATE) estimation in practice demands three properties simultaneously: heterogeneous effects $\tau(x)$, calibrated uncertainty over them, and robustness to the heavy tails that contaminate real outcome data. Meta-learners (K\"unzel et al., 2019) give (i); causal forests and BART give (i)-(ii) with Gaussian-tail assumptions; no widely used tool gives all three. We present Bayesian X-Learner, an X-Learner built on cross-fitted doubly robust pseudo-outcomes (Kennedy, 2020) with a full MCMC posterior over $\tau(x)$ via a Welsch redescending pseudo-likelihood. On Hill's IHDP benchmark the default configuration attains mean $\sqrt{\varepsilon_{\mathrm{PEHE}}} = 0.56$ on 5 replications (lowest mean; differences from S-/T-/X-learners, full-config Causal BART, and a causal forest baseline are not significant at $\alpha=0.05$, and rank ordering is unstable at 10 replications -- IHDP comparisons are competitive rather than dominant). On contaminated"whale"DGPs with up to 20-25% tail density, a one-flag extension (contamination_severity) that selects a Huber-$\delta$ nuisance loss per Huber's minimax-$\delta$ relation recovers RMSE $\approx 0.13$ with tight credible intervals (single-cross-fit 30-seed coverage 83% [Wilson 66%, 93%] at 20% density; modular-Bayes pooling with Bayesian-bootstrap nuisance draws restores nominal 95% coverage).